🚀 go-pugleaf

     -=> Timothy D. Goss spoke of "Re: splitting the group" <=-
    To All at 08-21-92  03:34
 TDG> Now.  Would anyone like to discuss some of the primitives necessary to
 TDG> make a really useful genetic algorithm?  I can't see how a 'good'
 TDG> critter can evolve without some powerful primitives to draw upon.  Any
 TDG> ideas?

I thought I posted something on this.  Anyway my thought is to
use "inheritance" combined with fractal geometry.  What I mean is
using a system of "weights" inherited from an ancestor these
weights each have an "equation" that's models a particular "state"
of behavior.

It's like a soccer ball on a field of play.  You have 20 players
each with their own bias.  10 want to score a goal to the right
10 to the left.  Each team of players vie for control of the ball.
In general it will depend on the strategies each player employees
on the moving the ball as to who wins.  however it also depends
on who has control of the ball!  SO let's say each of the ten
players have their own favorite strategy which will work
differently.  Each team struggles for control of the ball
meanwhile.

By inheriting these weights [that is the better ones] and
equations each succesive generation will have a better strategy
model to work with.  Besides by also adding NEW equations/weights
you could add mutation! :)

        Stephen Cyberman@Toz.Buffalo.NY.US
        Scripted at Fri  08-21-1992  14:00:47

... The more you run over a cat, the flatter it gets.
--- Blue Wave/QWK v2.10